Traditionally, when simulating or testing data systems, a proposed change to the system is installed and a series of tests, sometimes called regression tests are performed to determine if the data system with the change meets the design target and at the same time is compatible with previous versions of the data system. Regression testing tends to be exhaustive and binary. That is, a set of test cases are designed and developed. Each test case is performed and the results noted. Testing continues until either the first test fails or a catalog of failed tests is developed, at which time the testing is stopped and the data system is referred back to the development team.
In large scale distributed systems, changes made to portions of the system, as well as incremental changes made to each element of the system, may not be exhaustively testable, nor may the impact of a change be apparent from a given test case suite. Therefore, a system and method for monitoring and reacting to peer-to-peer network metrics is needed.